Wednesday, February 28, 2007

2007: Where Are The Clock Speeds?

The most significant CPU event in 2006 was, without doubt, Intel's release of Core 2 Duo. This will be followed this year by AMD's equally significant release of K10. With a major release from both companies the world of microprocessors would be rosy indeed . . . if not for the nagging question of clock speeds.

Before Core 2 Duo was actually released in mid 2006 there were announcements that Intel would release 3.2Ghz C2D's later in the year. There were even persistent rumors by Intel enthusiasts that 3.2Ghz chips would be available at launch. The latest roadmaps seem to indicate however that even 18 months later, Intel will still be stuck at 3.0Ghz. This however doesn't seem to have dented the enthusiasm of the Intel faithful who were certain that 3.2Ghz chips would be available before the end of 2006 and that at least 3.4Ghz chips would be available before the end of 2007. Most have now simply resorted to chattering about 45nm (Penryn). Supposedly, Intel 45nm is going to allow phenomenal speeds at low power draw. And, these eager souls are hoping fervently that Intel will somehow be able to release 45nm in Q3 07, presumably to compete with AMD's K10. However, Q4 07 looks to be the earliest for 45nm from Intel and there is as yet no indication of higher clock speeds. AMD, not to be outdone, seems equally determined to deliver sub 3.0Ghz clock speeds. AMD gives no indication of getting faster than 2.9Ghz with dual core K10 on 65nm and only 2.5Ghz with quad core K10 in 2007.

So, where are the clocks? Why are AMD and Intel unable to keep increasing clock speed with 65nm and push through 3.0Ghz? One consideration is that AMD has had a history of second wave clocks. For example, the highest clock achieved on 130nm exceeded the initial clock speed of 90nm. And, the high clock on 90nm exceeded the initial speed for 65nm. If this pattern is followed then it would be logical that AMD could reach a higher clock on 65nm by the time 45nm is released. Perhaps this will be the case. The picture is much less clear with Intel since Intel has almost no track record of one architecture on multiple processes. The only crossover has been Smithfield/Presler. With these two we see no evidence at all of second wave clocks. While Smithfield topped out at 3.2Ghz on 90nm, Presler launched at 3.6Ghz.

This has, of course, happened before. In late 2002 Intel was stuck at 3.0Ghz while AMD was stuck at 2.0Ghz. AMD did manage in early 2003 to bump the Barton K7 up to 2.2Ghz and, at about the same time, Intel bumped the Northwood P4 up to 3.2Ghz. Intel eventually managed 3.46Ghz with Northwood on 90nm with its Extreme Edition in 2004. However, it certainly looks now like the clock ceiling has returned with both Intel and AMD stuck below 3.0Ghz. I suppose AMD enthusiasts might point out that AMD did have increasing clocks and only recently hit 3.0Ghz with FX-74 and 6000+ (on 90nm). The problem is that we see no indication of clocks beyond 2.9Ghz on 65nm during all of 2007 and up to Q2 08. Strangely, none of the Intel enthusiasts seem to be discussing the lack of a 3.2Ghz Conroe. I suppose if someone can forget that Intel promised a 5.0Ghz Tejas in 2005 and then promised Whitefield with CSI in 2006 then forgetting the lack of a 3.2Ghz C2D clock would be relatively easy. So, ignoring Intel's 18 month span with no clock increases, I suppose it is possible that 45nm really is much better and that Intel could deliver clock rates as high as 3.6Ghz in 2008. However, nothing like this has been indicated yet. And, with a history of second wave clocks I suppose AMD could hit 3.1Ghz on 65nm by mid 2008.

This still does not answer the question though. Since K8 is currently hitting 3.0Ghz on 90nm one would assume that 3.2Ghz would be possible on 65nm. So, why is K10 only hitting 2.9Ghz? I suppose with AMD the problem is the process. Presumably AMD can't soon get over 2.9Ghz without going above 120 watts. Looking at history for AMD this probably matches the release of K8 in 2003. K8 Sledgehammer was released at 1.8Ghz which was roughly equal to the 2.2Ghz Barton K7. This could suggest that the initial 2.3Ghz Barcelona will be roughly equal to the 2.8Ghz Santa Rosa Opteron. The speeds would also parallel the introductory speeds of dual core X2 which were down from the fastest single cores. This also suggests that the later release of 2.5Ghz Barcelona will match the current speed of 3.0Ghz K8. This would put the Q2 08 speed of 2.6Ghz ahead of the current fastest K8. However, it appears that QFX will likewise be stuck at 2.5Ghz. This does seem a bit slow to compete with Kentsfield, however, QFX is a dual socket board. Given Opteron's current near equality with Xeon at the same clock it seems somewhat possible that a dual socket 2.5Ghz QFX could keep up with a dual socket 3.0Ghz Clovertown competitor. However, this is not entirely clear since a true competitor would not use FBDIMM but DDR2 like QFX. So, perhaps Intel could stay ahead of QFX.

And, why is Intel still stuck at 3.0Ghz? In spite of claims that Intel has superior process technology the conclusion would have to be the same, that Intel is unable to clock higher. And, there is no similar history for Intel as there is with AMD. There has simply never been a time when Intel was unable to increase clock speed for 18 months. This is a year and a half with no improvement on the process. Very strange. It is possible that the early results for 45nm were so promising that Intel stopped developing 65nm. This could be the case. I also know that some will claim that Intel has lots of headroom because their chips overclock well. Unfortunately, bulk silicon chips are more prone to failure than SOI chips so Intel has to maintain significantly more margin than AMD. Simply put, careful overclocking by enthusiasts is not comparable to the same speed with a stock heatsink and fan in a standard case in a house with no air conditioning and high ambient temperatures in the summer. Still, I'm sure this myth of headroom will stay alive until Intel moves to SOI as well. If you can conveniently forget the 5.0Ghz Tejas then presumably you can also forget that Dothan, Yonah, and Prescott all overclocked significantly higher than the speeds Intel released. I suppose if Intel fans are guilty of amnesia then AMD fans would be guilty of over-optimism. I've seen people insist that K10 will have a 40% increase in Integer IPC which is purely ridiculous; 20% is much closer to reality.

In terms of single socket in 2007, AMD's competitiveness is somewhat split. For dual core, this would be a 2.9Ghz Kuma versus a 2.93Ghz Conroe. This is a trivial 1% difference in clock, so probably even. However, for quad core this would be a 2.5Ghz Agena versus a 3.0Ghz Kentsfield. This is a not so trivial 20% difference in clock. I'm guessing Intel will stay ahead on this one although I suppose being competitive with the second fastest clock (2.6Ghz) is a lot better than having no quad cores at all. All in all this should be sufficient to halt any gains in share by Intel. Additional gains in share by AMD would then probably depend on other factors like price, availability, and chipset quality. So, it does appear that 2006/2007 mirrors 2002/2003 in several ways. Perhaps we will once again see standard clocks above 3.0Ghz in 2008.

Tuesday, February 20, 2007

Chipsets And Chess Pieces

Coming into 2007 the chipset market has shifted dramatically with AMD's acquisition of ATI. This not only makes AMD's position considerably stronger but will have effects on Intel's chipset direction as well.

While nVidia and ATI have been slugging it out in the discrete graphics market, Intel has enjoyed a fairly secure position in terms of desktop and mobile integrated chipsets. Intel's nearest integrated competitor was ATI, however ATI was not only a distant second but often depended on orders from Intel to have significant sales. Intel has enjoyed its ability to flex its muscle in this area as it did when it stopped purchases from VIA when they would not pay bus licensing fees. This has reduced VIA from being a strong chipset and graphics player to its current status as not much more than a supplier in the budget niche market. However, with AMD's purchase of ATI Intel's position has weakened dramatically.

In one stroke, ATI not only gains secure funding for R&D to compete with nVidia but it also gains early notification of strategic direction as well as close cooperation with AMD's CPU designers. This gives the ATI division a definite advantage over nVidia and puts ATI in a category that was the sole domain of Intel's chipset and graphics division. While most computer enthusiasts will see this as competition in the discrete area this is not the big fight. I do expect ATI to gain back some of the graphics share that it lost to nVidia however this won't be as dramatic as it will be for Intel. ATI and nVidia have been competing for a long time and are very used to this; the company that isn't used to this is Intel. Having lost its integrated factory supplied chipset monopoly, Intel can expect competition to greatly increase for corporate and commercial sales.

The commercial sales area is defined much less by performance and more by cost and reliability. OEM's want the systems to be easy to configure and maintain and as low of cost as possible. This is the stronghold of the integrated graphics chipset where the graphics chip is already attached to the motherboard. ATI's presence in this market just got a big boost since ATI chipsets can now be shipped with AMD's name and official endorsement clearly stenciled on each package. There is no doubt that AMD is deadly serious about this market and is diving in with a headstart over Intel. The reason for this headstart has to do with cost.

Every chip on a motherboard costs something, and every chip connects to other chips and makes motherboard design more complex. The best way to reduce cost is to reduce the chip count. AMD already has a huge leg up on Intel because AMD motherboards don't need a Northbridge chip as all Intel motherboards do. This reduces both cost and complexity. AMD is continuing down the chip count reduction road with its plan to put the graphics chip (GPU) on the same die as the CPU. This sort of thing has been done in the past as when the MMU and FPU were integrated on the CPU die in the same fashion. AMD has went further by including the memory controller as well. So, in 2008 there should be plenty of smiles from motherboard designers who can leave off the Northbridge and GPU and design a simpler, cheaper motherboard. These smiles should be passed along to commercial OEM's and finally to commercial customers. Simple, inexpensive, easy to validate and configure. This is exactly what the market needs.

Intel however seems to have been caught off guard in this area. Presumably, it has been financing the added cost of the Northbridge with the money it has been getting from bus licenses. However, it may now not have this revenue from ATI. Intel had talked about putting the GPU inside the Northbridge which would indeed reduce the chip count. It appears that Intel has done this with the iG965. However, one has to wonder if this move is very bright when presumably the Northbridge will disappear in 2009 when Intel releases the new X86 socket with CSI and the memory controller included on the CPU die. In other words, when Intel tries to follow AMD by dropping the Northbridge then whatever work it has put into having the GPU on the Northbridge will get tossed out. This certainly isn't additive development but Intel probably has the money to waste. However, since the iG965 still has to connect to memory and the FSB of the CPU even adding the GPU doesn't reduce the complexity as much as it does on AMD motherboards. Also, it has been suggested that intel will need a switching hub to connect CSI among the processors so this would be adding a new chip on the motherboard. This too creates doubt that Intel will be able to match AMD in terms of reducing chip count and motherboard cost.

Intel may try to follow AMD by putting the GPU on the CPU die however I'm certain it will be at least a year behind AMD. Whereas AMD's K8 uses a crossbar to communicate among the HyperTransport links, the Memory Controller, and the CPU Cache, Intel's chips have no such functionality. AMD can easily hook up a GPU by adding another port to the crossbar. However, since Intel chips have no crossbar they will need to invent some method of communication. In the past this was done via Northbridge translation between the GPU and FSB of the CPU. Now that the FSB will be dropped along with the Northbridge Intel will need something entirely new. I suppose the quick and dirty approach would be to use CSI. This could work but CSI is a long way from HyperTransport's track record of robust, reliable communication. In fact, it is perhaps a somewhat embarassing secret that some 3rd party Intel motherboards use AMD's HyperTransport to talk between the North and Southbridge chips. In the BIOS this is referred to by the original name, Lightning Data Transport or LDT.

So, while Intel currently has the stronger position in terms of processors they will have to scramble to keep from having the chipset rug pulled out from under them in a determined and focused assault by AMD. AMD's drive to reduce cost and complexity by reducing chip count will be hard to match, not only by Intel but by other 3rd party chipset makers like VIA. This does leave the question of whether low end computers could be reduced in cost even further. And, since any further reductions in low end system cost along with potential increases in power could overlap with the set-top market one has to wonder if XBOX and Playstation will still be around in 2009. However, I have little doubt that by 2009 ATI's share of the discrete graphics market will be much closer to nVidia's than it is now and ATI's share of the desktop and mobile integrated markets will be much closer to Intel's.

Wednesday, February 07, 2007

Intel's 2007 Processors: Mist In The Morning Sun

After more than two months of speculation about Intel's 2007 processor offerings it seems that the mist is finally clearing. However, what is left behind is much less than was suggested. It is now clear that the speculation greatly overestimated what Intel would be capable of delivering.

Most of the talk about Intel's 2007 processor offerings grew out of two sources: the VR-Zone roadmaps and glowing reports about Intel's 45nm process. According to VR-Zone, Intel was to release the 45nm quad core Yorkfield chips in Q3 07 along with the 45nm dual core Wolfdale. As seen in the chart, Yorkfield would clock from 3.46ghz to as high as 3.73Ghz. Likewise, Wolfdale would clock from 3.5Ghz to as high as 4.0Ghz. Clock speeds that high would have surpassed anything AMD could have offered and given Intel a tremendous lead. Any scepticism about these clock numbers was quickly brushed aside as people waxed almost poetic about Intel's 45nm process. Supposedly it would have allowed huge increases in clock speed while also drawing less power. The typically stated cornerstone of Intel's 45nm process was high-K, or at least, this was the only item that Intel had that AMD did not. Adding to this, AMD's 65nm Brisbane chips were not readily available in the channel in Q4 06. When they did become available in the channel they were at lower clocks than the existing 90nm chips. So, the chatter immediately moved to talk about problems with AMD's 65nm process and even speculation that Barcelona wouldn't arrive until 2008. Not surprisingly, when Intel recently announced that it had actual working chips with high-K, the press reports were overflowing with praise. The second half of 2007 looked to many like it would be a landslide success for Intel with performance even further ahead of Barcelona than C2D was ahead of K8.

However, in the latest roadmaps, all of that is gone; Intel's amazing 45nm speed has completely vanished. The highest clocking Yorkfield is now only 2.4Ghz and Wolfdale tops out at 3.0Ghz. Worse still, even with these reduced speeds, 45nm has been pushed back to Q1 08. And, adding insult to injury, even Anandtech now agrees that AMD's Barcelona quad core will begin production in Q3 07 at 2.5Ghz. This makes sense for a lot of reaons however it seems that history and common sense are often casulties of processor enthusiasm. For example, when Core 2 Duo arrived, it probably seemed that Intel's Midas Touch had returned. The problem with this view however was that C2D's improvement was not unique. AMD had a similar increase in IPC with K8 and it now appears that Barcelona will have a similar increase in IPC over K8. SSE was the only area where C2D saw phenomenal increases in speed and this was done by doubling the width of all the necessary hardware. However, Barcelona's FP/SSE hardware will receive a similar stretch in width. If Barcelona is any slower than C2D at the same clock, it won't be by much. Rather than just being a simple upgrade, it has now been indicated that K8 was completely resurveyed and many improvements were made all over the architecture including branch speculation, out of order loading, stack operations, bus access scheduling, prefetch, and instruction decoding. Barcelona is therefore obviously a new core in the same way that K8 was. Presumably, Barcelona would then be the K9 or K10 core but AMD has not indicated this yet.

So, 2007 is now shaping up to be pretty much even in terms of performance between Intel and AMD. And, 2008 shouldn't be much different since IBM/AMD also have high-K and will use it on 45nm just as Intel will. The talk about what Intel would offer in 2007 looked like giants in the mist. And, now that the mist is clearing we can see that Intel's processors don't stand any taller than AMD's. It looks like we are going to have some very nice processors from both AMD and Intel in late 2007 and early 2008. It looks like both will have good Integer and SSE performance with very similar IPC and power consumption. This should be great for customers of both companies. Nevertheless, this still seems to be unacceptable to a lot of enthusiasts. Perhaps this is why the talk now seems to be about AMD's supposed poor finances and speculation about bankruptcy. I'm sorry but AMD is not going to go bankrupt while introducing a new architecture, new mobile chipset, and being able to compete better in the commercial segment with its desktop, integrated chipsets.

An interesting final note is how popular this blog is with Intel employees. For example, AMD employees are no more likely to read my blog than employees of Pratt & Whitney or General Electric. However, I get 20 times as much traffic from Intel employees. In fact, Intel employees are the single largest group of readers. I've seen this blog dismissed a number times as blatantly pro-AMD FUD by various people (such as the Tommies at ForumZ). However, I do have to wonder then, if this blog contains so little valid information, why readers from Intel keep coming back.